A super computer to predict revolutions
It does not yet exist. But study suggests it will someday be possible
On December 6, 1941, the Foreign Broadcast Information Service (FBIS), a service responsible for listening to foreign radio stations, set up by the American intelligence community and one of the first experiments of what is now called "open source" intelligence, delivered its first report: an analysis of the feeling of the Japanese media towards the United States. The report noted that Japanese radio stations had seen their criticism of the United States soar and had ceased their calls for peace. The next day, Pearl Harbor was bombed.
Obviously, no listening to the media could have revealed when and where the attack would take place (that's why we have spies), but it is likely that with a better understanding of these signals preparatory to an attack , the American forces would not have been so surprised. 70 years later, a computer specialist even believes that a slightly more ambitious version of this same type of information analysis will soon be able to predict social disturbances and conflicts - like the recent revolutions in the Arab world - with a degree remarkable accuracy.
Kalev Leetaru, Deputy Director for Text and Digital Analysis at the Institute for Computer Science Applied to the Arts, Letters and Human Sciences at the University of Illinois is one of the leading researchers in an emerging field, prediction conflicts. In a study published this month in an online technology journal, the articles of which are evaluated anonymously by a panel of specialists, First Monday, Leetaru claims that "quantitative analysis of large sequences of text can give new insights into the functioning of society. "
Leetaru's study transports recent economic research and examines how analysis of the tone of news and social media can predict certain economic events. A recent study, for example, has shown that analysis of general sentiment on Twitter can anticipate movements in the Dow Jones index. Leetaru was curious to know if this same type of analysis could also predict social events.
Sentiment analysis software
Leetaru has used several databases of news articles from the past 30 years, including the "Summary of World Broadcasts" - English translations of foreign radios produced by the British equivalent of FBIS - the complete digital archives of the New York Times , and an analysis of online news sites to create a database of almost 100 million news articles dating back to 1979.
Then he got this raw material into one of the most powerful supercomputers in the world, the University of Tennessee's "Nautilus", and he started looking for correlations.
(...)
Suite and source: http://www.slate.fr/story/44315/super-o ... evolutions